I recently took two Databricks certification exams: Data Engineer Professional and, last night, Generative AI Engineer Associate.
For my first exam, I was very stressed because I had no previous Databricks certification and did not fully know what to expect.
The first 10 questions felt difficult, and I was already thinking:
"Maybe I should have taken the Associate exam first."
But later I found easier questions, increased my speed, reached the end, and used the remaining time to review marked questions. I kept fighting until the timer ended.
My second exam felt different.
Before the GenAI Associate exam, I was much more confident. But once it started, I saw many long scenario-based questions and immediately felt strong time pressure.
My exam had 56 questions in 90 minutes, while the official scored-question count is 45.
That means roughly 25% more questions with 0% more time.
This is why I believe unscored questions are not truly neutral.
They may not directly affect the score, but they still consume time, attention, and mental energy. Every minute spent on an unscored question is a minute that cannot be spent on a scored one.In the final minutes, I still had three questions to finish and had to answer them very quickly, partly based on educated guesses.
I understand that Databricks needs experimental questions to improve future exams, but I think the current approach creates a fairness issue. A better approach could be to make experimental questions optional, with proportional extra time and perhaps an incentive such as a future exam discount.
My main feedback is simple:
Unscored questions can still negatively affect a candidateโs scored performance when the total exam time stays fixed.
And one final lesson from both exams: never give up because the beginning feels difficult or because the clock is against you.
Fortunately, I passed the exams.
